Clinical Utility and Validation of the Krakow DCM Risk Score—A Prognostic Model Dedicated to Dilated Cardiomyopathy

Background: One of the most common causes of heart failure is dilated cardiomyopathy (DCM). In DCM, the mortality risk is high and reaches approximately 20% in 5 years. A patient’s prognosis should be established for appropriate HF management. However, so far, no validated tools have been available for the DCM population. Methods: The study population consisted of 735 DCM patients: 406 from the derivation cohort (previously described) and 329 from the validation cohort (from 2009 to 2020, with outcome data after a mean of 42 months). For each DCM patient, the individual mortality risk was calculated based on the Krakow DCM Risk Score. Results: During follow-up, 49 (15%) patients of the validation cohort died. They had shown significantly higher calculated 1-to-5-year mortality risks. The Krakow DCM Risk Score yielded good discrimination in terms of overall mortality risk, with an AUC of 0.704–0.765. Based on a 2-year mortality risk, patients were divided into non-high (≤6%) and high (>6%) mortality risk groups. The observed mortality rates were 8.3% (n = 44) vs. 42.6% (n = 75), respectively (HR 3.37; 95%CI 1.88–6.05; p < 0.0001). Conclusions: The Krakow DCM Risk Score was found to have good predictive accuracy. The 2-year mortality risk > 6% has good discrimination for the identification of high-risk patients and can be applied in everyday practice.


Introduction
Dilated cardiomyopathy (DCM) is the commonest indication for heart transplantation and the third most common cause of heart failure (HF) [1][2][3][4][5]. It is characterized by left ventricular (LV) systolic dysfunction and LV enlargement in the absence of significant coronary artery disease and abnormal loading conditions [1][2][3][4][5]. Over the past few decades, the aetiology and natural history of DCM have been thoroughly elucidated, demonstrating that various aetiologies causing LV dysfunction may manifest with the same clinical phenotype as DCM.
Providing accurate prognoses in HF can pose numerous challenges. So far, many scales, including BCN Bio-HF, CHARM, EMPHASIS, HFSS, MAGGIC, MUSIC and SHFM, all dedicated to the general HF cohort, have been developed, with a diagnostic accuracy of between 60% and 80% [6][7][8][9][10][11][12][13][14][15]. However, most were created 10-20 years ago, before the global implementation of HF modifying therapies. These treatments substantially diminish the applicability of these scales to current HF populations, especially given the fact that they have not been validated for any subgroup of HF, including DCM. The validation of the scales is of utmost importance to DCM patients, as they differ substantially from other types of HF patients in terms of their younger age and their presentation of fewer comorbidities, which leads to an overall lower mortality rate [3,16].
At present, there are only two prognostic scales dedicated to DCM: Miura et al. and the Krakow DCM Risk Score [12,17,18]. Although the first is a simple numerical score based on five parameters (and as such is easy to calculate), its prognostic value is questionable, especially since it was developed before current HF therapies were introduced [18]. The second one is a linear scale that has performed very well in bootstrapping, and despite its complexity, an online tool is now available. Nevertheless, it has not as yet been externally validated [17,18].
Therefore, the aim of this work is to externally validate the Krakow DCM Risk Score and to establish a cut-off point for high-risk DCM patients.

Clinical Follow-Up and Endpoint Definition
As in the original derivation study, the endpoint was all-cause mortality. Between May and September 2021, information on the status of the patients was collected through publicly available databases, medical records and telephone contact.

Statistical Analysis
All parameters are presented as mean ± standard deviation (SD) or counts and percentages. The continuous parameters were tested for their normal distribution with the Shapiro-Wilk test. Comparisons of quantitative variables were conducted with t-tests or the Mann-Whitney test for data with and without normal distribution; the Chi-square test was performed in the case of qualitative parameters. Areas under the receiver operating curve (AUC) were calculated to assess the accuracy of the Krakow DCM Risk Score for the prediction of 1-, 2-, 3-, 4-and 5-year mortality. Kaplan-Meier analyses were performed for the calculation of observed mortality and the log-rank test for the comparisons of mortality rates. Results were considered statistically significant when their p-value was <0.05. The Statistica package, version 13.0 (StatSoft, TIBCO Software Inc., Palo Alto, CA, USA), was used for the statistical analysis.

Baseline Characteristics
During a follow-up of mean 41.6 ± 29.3 months, 49 (15%) patients of the validation cohort died: 41 (85%) stemming from cardiovascular causes (38 patients-due to HF worsening, 3 patients-SCD), and eight patients died as a result of neoplasms. In terms of procedures, 12 patients underwent left ventricle assistant device (LVAD) implantations and six patients heart transplants (HTX); one patient received both procedures.
Deceased were older, more symptomatic, had more severe LV and right ventricular (RV) remodelling with worse LV and RV systolic function, more often had significant tricuspid regurgitation, anaemia and chronic kidney disease, higher levels of N-terminal prob-type natriuretic peptide (NT-proBNP), and they required higher loop diuretics dosages (Table 1). HF modifying therapies, such as beta blockers and renin-angiotensin-aldosterone system inhibitors, were more commonly used in both groups, whereas digoxin, higher doses of loop diuretics and CRT were more prevalent among deceased.

Performance of Krakow DCM Risk Score
Calculated mortality risks, based on the Krakow DCM Risk Score, significantly differed between alive and deceased patients ( Table 2). The model under analysis yielded good discrimination in terms of overall 1-, 2-, 3-, 4-and 5-year mortality with an AUC of 0.704-0.765 ( Figure 1). The validation cohort differed significantly from the derivation cohort in terms of age, HF duration and symptoms, comorbidities (obesity-body mass index, dyslipidaemia, anaemia and chronic kidney disease), heart rate, NT-proBNP and their required loop diuretics dosage (Table S1, Appendix B). However, they did not differ in terms of echocardiographic findings and mortality rates (p = 0.97) ( Figure S1).

Analysis)
All Alive Deceased p-Value AUC-ROC

High Mortality Risk DCM Patients
Although the mean observation period for the whole DCM cohort (derivation and validation cohorts) was 45 months, only 60% (n = 425) of the entire population had followup longer than 3 years, 50% (n = 356) longer than 4 years and 39% (n = 273) longer than 5 years. Therefore, high mortality risk was assessed based on 2-year mortality risk as calculated by the Krakow DCM Risk Score in those patients (n = 735) with available 2-year follow-up data.

High Mortality Risk DCM Patients
Although the mean observation period for the whole DCM cohort (derivation and validation cohorts) was 45 months, only 60% (n = 425) of the entire population had followup longer than 3 years, 50% (n = 356) longer than 4 years and 39% (n = 273) longer than 5 years. Therefore, high mortality risk was assessed based on 2-year mortality risk as calculated by the Krakow DCM Risk Score in those patients (n = 735) with available 2-year follow-up data.

Discussion
The study findings can be summarised as follows: the Krakow DCM Risk Score yielded adequate discrimination in terms of overall mortality in the DCM population. The cut-off point of 6% for a 2-year mortality risk displayed good discrimination for high mortality risk DCM patients.

Prognostic Models in DCM
The observed mortality in DCM patients is high, and over the course of 5 years reaches approximately 20% but varies among different studies that have been carried out [12,19,[23][24][25][26]. However, due to its unique features (such as its occurrence at a young age, LV reverse remodelling and fewer comorbidities), risk stratification in DCM cannot be accurately performed using unspecific models developed for broad HF populations [27,28].
Although numerous prognostic parameters have been established in DCM, including HF symptoms severity (mostly assessed by semi-quantitative NYHA class), LV and RV systolic function and size, comorbidities (e.g., diabetes mellitus, anaemia, chronic kidney disease), cardiac fibrosis, or ventricular arrhythmias, heir clinical meaning in isolation has limited value for more thorough-going risk stratification [12,[28][29][30][31][32]. Consequently, until recently, there were no tools in existence for accurate mortality risk stratification in DCM. So far, two prognostic models dedicated to DCM patients exist: (1) the Miura et al. score and (2) the Krakow DCM Risk Score [12,19]. Miura et al. is a numerical score based on a Japanese national DCM survey from the 1990s, which calculates the 5-year mortality risk

Discussion
The study findings can be summarised as follows: the Krakow DCM Risk Score yielded adequate discrimination in terms of overall mortality in the DCM population. The cut-off point of 6% for a 2-year mortality risk displayed good discrimination for high mortality risk DCM patients.

Prognostic Models in DCM
The observed mortality in DCM patients is high, and over the course of 5 years reaches approximately 20% but varies among different studies that have been carried out [12,19,[23][24][25][26]. However, due to its unique features (such as its occurrence at a young age, LV reverse remodelling and fewer comorbidities), risk stratification in DCM cannot be accurately performed using unspecific models developed for broad HF populations [27,28].
Although numerous prognostic parameters have been established in DCM, including HF symptoms severity (mostly assessed by semi-quantitative NYHA class), LV and RV systolic function and size, comorbidities (e.g., diabetes mellitus, anaemia, chronic kidney disease), cardiac fibrosis, or ventricular arrhythmias, heir clinical meaning in isolation has limited value for more thorough-going risk stratification [12,[28][29][30][31][32]. Consequently, until recently, there were no tools in existence for accurate mortality risk stratification in DCM. So far, two prognostic models dedicated to DCM patients exist: (1) the Miura et al. score and (2) the Krakow DCM Risk Score [12,19]. Miura et al. is a numerical score based on a Japanese national DCM survey from the 1990s, which calculates the 5-year mortality risk based on just five parameters: sex, age, NYHA class, LV diameter and LVEF [12]. Although the calculation is straightforward, its overall performance is far from satisfactory [19]. Moreover, apart from our own external validation in a contemporary European cohort, the Miura score has never been validated. On the other hand, the Krakow DCM Risk Score, based on 406 DCM patients from 2010 to 2019, is a linear model that allows for the calculation of the individual mortality risk at any given time, preferably between 1 and 5 years [17,19]. To facilitate the use of the proposed model, an online calculator has been created, which is available on the Heart Failure Association of the Polish Cardiac Society official webpage (Figure 3). J. Pers. Med. 2022, 12, x FOR PEER REVIEW 7 of 11 based on just five parameters: sex, age, NYHA class, LV diameter and LVEF [12]. Although the calculation is straightforward, its overall performance is far from satisfactory [19]. Moreover, apart from our own external validation in a contemporary European cohort, the Miura score has never been validated. On the other hand, the Krakow DCM Risk Score, based on 406 DCM patients from 2010 to 2019, is a linear model that allows for the calculation of the individual mortality risk at any given time, preferably between 1 and 5 years [17,19]. To facilitate the use of the proposed model, an online calculator has been created, which is available on the Heart Failure Association of the Polish Cardiac Society official webpage (Figure 3).

Krakow DCM Risk Score Performance
According to the analysis presented here, the Krakow DCM Risk Score shows adequate performance in external validation, with an accuracy of over 70% [14]. This precision is comparable to the most widely available tools currently in use, such as the GRACE risk score 2.0 for mortality outcomes in acute cardiac syndrome, the HCM Risk-SCD score for sudden cardiac death (SCD) in hypertrophic cardiomyopathy (HCM), or CHA2DS2-VASc for stroke in atrial fibrillation, and is at least similar to prognostic scores in general HF cohorts, including the Heart Failure Survival Score (HFSS), the Seattle Heart Failure Model (SHFM) and the Meta-Analysis Global Group in Chronic Heart Failure (MUSIC) [14,15,[33][34][35][36]. The Krakow DCM Risk Score provided good discrimination for at least 7 out of 10 patients. Taking into account the significant differences between the derivation and validation cohorts, including age, HF symptoms and NT-proBNP level, their outcomes were poor with similar 5-year mortality of over 20%. Moreover, the high accuracy of the Krakow DCM Risk Score, despite the diversity of the DCM cohort's understudy, strengthens its value in various DCM cohorts.

Identification of High Mortality DCM Patients
Accurate prediction of long-term outcomes, including mortality, is a cornerstone of comprehensive HF management. Although the overall prognosis in DCM is poor, patients' individual prognoses may be highly variable. Therefore, the development of an accurate prognostic risk model has the potential for more comprehensive and tailored management. Quantifying patients' survival predictions based on their overall risk profile can

Krakow DCM Risk Score Performance
According to the analysis presented here, the Krakow DCM Risk Score shows adequate performance in external validation, with an accuracy of over 70% [14]. This precision is comparable to the most widely available tools currently in use, such as the GRACE risk score 2.0 for mortality outcomes in acute cardiac syndrome, the HCM Risk-SCD score for sudden cardiac death (SCD) in hypertrophic cardiomyopathy (HCM), or CHA 2 DS 2 -VASc for stroke in atrial fibrillation, and is at least similar to prognostic scores in general HF cohorts, including the Heart Failure Survival Score (HFSS), the Seattle Heart Failure Model (SHFM) and the Meta-Analysis Global Group in Chronic Heart Failure (MUSIC) [14,15,[33][34][35][36]. The Krakow DCM Risk Score provided good discrimination for at least 7 out of 10 patients. Taking into account the significant differences between the derivation and validation cohorts, including age, HF symptoms and NT-proBNP level, their outcomes were poor with similar 5-year mortality of over 20%. Moreover, the high accuracy of the Krakow DCM Risk Score, despite the diversity of the DCM cohort's understudy, strengthens its value in various DCM cohorts.

Identification of High Mortality DCM Patients
Accurate prediction of long-term outcomes, including mortality, is a cornerstone of comprehensive HF management. Although the overall prognosis in DCM is poor, patients' individual prognoses may be highly variable. Therefore, the development of an accurate prognostic risk model has the potential for more comprehensive and tailored management. Quantifying patients' survival predictions based on their overall risk profile can help identify those in need of more concentrated monitoring and more intensive HF therapy. Additional potential use of the model includes educating patients on the significant value of HF medications. When altering their implementation in the online calculator, patients can be presented with a higher mortality risk without proper pharmacotherapy. Moreover, patients with high mortality risk should be earlier listed for cardiac transplantation or counselled about end-of-life issues.
As stated above, the identification of patients with high mortality risk is crucial in everyday practice in HF and DCM patients. We recognized the high mortality risk as a 2-year mortality risk estimated above 6.0% (75th percentile). The accuracy of this cut-off point is high, and patients with a calculated risk of >6.0% had a five-times higher mortality risk during follow-up. Therefore, it can be used in everyday practice for HF therapy qualification, especially in terms of invasive procedures.

Limitations
Although the size of the study population is large, with over 700 DCM patients, taking into account DCM epidemiology, this is still a one-country retrospective analysis. The mean observation period for the whole DCM cohort was 45 months; however, only half of the population had follow-up data on 4-year mortality. Therefore, to make the model more accurate, high mortality risk was assessed based on observations over the course of 2 years. Only 19% of patients had been treated with angiotensin receptor-neprilysin inhibitors; however, sacubitril/valsartan was only available for the second half of the study. Moreover, the Krakow DCM Risk Score does not include HTX and LVAD implantation as outcomes; however, overall mortality is the definite endpoint. Although the Krakow DCM Risk Score has a complicated linear model, it includes the most known DCM prognostic parameters, and its discrimination is good. Therefore, to facilitate the use of the model, an online calculator has been created and made widely available.

Conclusions
The overall mortality risk in the DCM population is high and reaches 23% during 5-year follow-up. A DCM-dedicated prognostic model, namely the Krakow DCM Risk Score, was found to have good predictive accuracy. The 2-year mortality risk of over 6.0% has good discrimination for the identification of high-risk patients and can be used in everyday practice.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/jpm12020236/s1, Table S1: Comparison of the output and validation cohorts. Table S2: Comparison of baseline characteristics of patients with high and non-high mortality risk, calculated as the 2-year mortality risk according to the Krakow DCM Risk Score> or ≤6.0%. Figure S1. Comparison of mortality rates between derivation and validation cohorts based on Kaplan-Meier estimates.  Informed Consent Statement: Patient consent was waived due to the retrospective character of the study.
Data Availability Statement: Data available on request due to privacy restrictions. The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.